Predicting direction shifts on Canadian–US exchange rates with artificial neural networks

Abstract

The paper presents a variety of neural network models applied to Canadian-US exchange rate data. Networks such as backpropagation, modular, radial basis functions, linear vector quantization, fuzzy ARTMAP, and genetic reinforcement learning are examined. The purpose is to compare the performance of these networks for predicting direction (sign change) shifts in daily returns. For this classification problem, the neural nets proved superior to the naïve model, and most of the neural nets were slightly superior to the logistic model. Using multiple previous days' returns as inputs to train and test the backpropagation and logistic models resulted in no increased classification accuracy. The models were not able to detect a systematic affect of previous days' returns up to fifteen days prior to the prediction day that would increase model performance.

Using Artificial Neural Networks to forecast Exchange Rate, including VAR‐VECM residual analysis and prediction linear combination

“…Econometric models have been widely applied for this task; however, they have also been heavily criticized as the majority are linear and work under assumptions that restrict them. Thus, the use of artificial intelligence techniques has spread, such as Artificial Neural Networks (ANN), which are able to model the non‐linear behavior of time series through their learning, adaptability, and training properties, without needing to know the data structure in advance, favoring models focused on forecasting ((Chen & Leung, ); (Davis, Episcopos, & Wettimuny, ); (Kodogiannis & Lolis, ); (Yao & Tan, )).Despite the discrepancies between different studies on exchange rate linearity, the literature shows a greater interest in non‐linear models to analyze the behavior of this time series ((Clements & Lan, ); (Yu, Wang, & Lai, ); (Leung, Chen, & Daouk, )). The goal of this study is to verify that there is an improvement in the forecasts of exchange rate returns when using embedded models as opposed to using only econometric or artificial intelligence models, all in a rolling windows frame.…”

Linear Relationship Between the Aud/Usd Exchange Rate and the Respective Stock Market Indices: A Computational Finance Perspective

“…Davis et al . () compared a set of ANN models in predicting directional shift for the Canadian dollar/USD exchange rate and noted that the ANN models outperformed the traditional models. Kamruzzaman and Sarker () demonstrated that an ANN‐based model performed better than the classical ARIMA model in predicting exchange rates of the AUD against six other currencies.…”

“…With the advancement of computational technologies, there have been recent efforts to employ intelligent system techniques like artificial neural network (ANN) and support vector machine in varied financial issues, including exchange rate prediction (e.g. Davis et al, 2001;Gradojevic and Yang, 2006;Emam, 2008). These computational intelligence techniques can realize complex mathematical relationships between predictors (i.e.…”

“…Another early study (Hu and Tsoukalas, 1999) examined the performance of four models (including the ANN) for European Monetary System exchange rates and pointed out an exponential general autoregressive conditional heteroskedastic (EGARCH) model as the best predictor. Davis et al (2001) compared a set of ANN models in predicting directional shift for the Canadian dollar/USD exchange rate and noted that the ANN models outperformed the traditional models. demonstrated that an ANN-based model performed better than the classical ARIMA model in predicting exchange rates of the AUD against six other currencies.…”

Continuous ant colony optimization algorithms in a support vector regression based financial forecasting model

“…Recently, accompanying the mature nonlinear mapping capabilities and data processing characteristics, artificial neural networks (ANNs) have received much increasing attention in financial forecasting. Many researchers had applied ANNs concepts to construct appropriate forecasting models to implement forecasting works, such as original ANNs models [11][12][13][14][15][16][17][18], hybrid models of fuzzy logic [19], multi-layer feed-forward network [20,21], and general regression neural networks (GRNN) [22], etc. All of these studies had demonstrated that ANNs-based models outperform the FEA1 forecasting models.…”